Biopharmaceutical Process Development & Manufacturing Final Project
1. Final Project
~Team #5 on Topic #4~
Mitigation of mAb Aggregation in Drug Substance Through Downstream
Process Optimization
Students: Will Carroll, Jonah Spielman, Kenneth Daniel, Victoria Null
Table of Contents
Problem Description 1
Experimental Overview & Strategy 2
Phase 1: Rank mAb Types for Likelihood of Aggregation (5 Working Days) 2
Phase 2: Assessment of pH Dependence for Aggregation Rates (30 Working Days) 2
Phase 3: Assess Process Performance with Respect to the Elution and Hold pH (55 Working
Days) 3
Phase 4: Vary Excipient Type/Buffer Additives and Concentration (30 Working Days) 3
Experimental Details 4
Phase 1 4
Phase 2 4
Phase 3 7
Phase 4 9
References 12
We pledge on our honor that we have not given or received any unauthorized assistance on this
assignment.
2. 1
Problem Description
There are often multiple sources of aggregation during downstream purification. Protein
aggregation decreases the safety, efficacy, and quality of mAbs. Currently, there is no consensus
on the maximum allowable aggregate levels in protein-based pharmaceutical products because
some proteins may be largely stable and safe despite certain levels of aggregates, while for other
proteins very small changes in aggregate levels may significantly affect protein stability and
efficacy. As such, limits for soluble protein aggregates have to be set on a case-by-case basis as
there are no regulatory limits for aggregates in biotherapeutic preparations. In the absence of
information on clinical relevance and process control, many specifications are instituted with
narrower than necessary acceptance ranges based on manufacturing experience. “The only formal
limit in USP 788 is on limiting sub-visible particles, which says that “Solutions for injection must
be clear and practically free from particles”” (Vázquez-Rey & Lang, 2011).
The issue of protein aggregation is important to address for Phase 1 clinical trials because
the goal of these trials is to find the optimal dose with the fewest side effects. Because aggregates
often have negative side effects, it is crucial that the amount of aggregates is minimized in the
downstream process, thus leading to a safe product for the consumer. This report will focus on the
amount of aggregates formed during Protein A Chromatography.
Protein A Chromatography is the most common mAb capture step that is used in the
downstream process. However, this technique requires the use of pHs around 3 to 4. At low pHs,
the mAb has the potential to undergo structural changes, thus leading to aggregation (Vázquez-
Rey & Lang, 2011). It is important to note that Protein A Chromatography does not remove
aggregates. This is due to the interactions of histidyl residues in Protein A and the binding site of
the mAb (Vázquez-Rey & Lang, 2011). At low pHs, both residues have positive charges. This
leads to the two residues repelling each other and the mAb eluting. As aggregates still contain the
histidyl residue, they are also eluted. Therefore, the challenge of minimizing aggregation formation
during Protein A Chromatography lies within stabilizing the protein at low pHs and finding a
balance between using a strong base and a large amount of base at a lower concentration (Vázquez-
Rey & Lang, 2011).
It is estimated that our proposed experiments will be completed within three months using
two scientists working eight hours per day. This amounts to 60 working days per scientist for a
total of 120 working days, during which time four project phases will be completed.
3. 2
Experimental Overview & Strategy
Phase 1: Rank mAb Types for Likelihood of Aggregation (5 Working Days)
The purpose of using TANGO, which determines aggregation propensity, is to predict
which mAbs are aggregation-prone (Kant, et al., 2017). This experiment will be the first phase of
analysis, since the predictions will be compared to those from the later experiments. This
experiment will not influence the mAbs chosen to be studied, however, it will provide valuable
information for future research and development. Additionally, it is assumed that the primary
sequence of all mAbs tested is available.
Phase 2: Assessment of pH Dependence for Aggregation Rates (30 Working Days)
Low pH methods of purification often lead to increased instances of aggregation among
mAb samples (Vázquez-Rey & Lang, 2011). Due to the many mechanisms of aggregation, the
predominant cause of aggregation remains unclear. In literature, it was hypothesized that a low pH
contributed to an abundance of negative charges that favorably contributes to repulsive interactions
that prevent aggregation, a condition of high colloidal stability (Wälchli, R, 2019). Instead, it was
thought that low pH induces protein denaturation. Typically, denaturing has involved partial
unfolding that exposes hydrophobic residues in the mAb (Vázquez-Rey & Lang, 2011). When this
denatured protein is returned to a neutral pH and the repulsive electrostatic forces are reduced,
aggregation can occur readily (Wälchli, R, 2019). This phase of the experiment will focus on the
pH-dependent effects of protein aggregation. This information will help to uncover a greater
understanding of Protein A Chromatography, Low pH Viral Inactivation, and their effects on mAb
aggregation. To consider the mechanism of aggregation proposed by Wälchli, mAb aggregation
will be evaluated both during a low pH hold and upon titration to a pH of 7. The low pH models
the Protein A elution and viral inactivation while the concluding neutral condition is the likely feed
to downstream polishing. The pH dependence of aggregation will be tested by the variance of the
low pH hold condition. SEC will be used to assess aggregation at neutral pH and 1-
anilinonaphthalene-8-sulfonate (ANS) fluorescence will be used to assess degradation at low pH
(Wälchli, R, 2019). The aggregation values at low pH would then be compared to the result from
neutral pH.
4. 3
Phase 3: Assess Process Performance with Respect to the Elution and Hold pH (55 Working
Days)
While phase 2 provides a clear understanding of the pH dependence of mAb degradation
rates, these results do not indicate the performance of Protein A chromatography or the low-pH
viral inactivation. The predominant roles of Protein A chromatography include, among other
functions, the removal of Host Cell (HC) proteins and DNA. Meanwhile, viral inactivation aims
to destroy any viral particles and bacterial contaminants. Given the diversity of potential
contaminants, designing an experimental approach requires substantial restrictions in the interest
of time. All tests are standardized and restricted to a hypothetical condition. Notably, it is assumed
that all mAb stock solutions are essentially pure, lacking both HC materials and viral contaminants.
Hence, all analyses will be restricted to the evaluation of (1) the mAb %
purity/aggregates/fragments in correspondence with Phase #2, (2) the HC-DNA presence in the
protein A eluent, and (3) the presence of select viral particles spiked into the mAb stock. These
observations are made with respect to a set eluent and hold pH obtained with the same stock buffers
tested in phase 2’s analysis. The effects of surfactants and excipients are not of interest in this
study and their presence is not varied.
Phase 4: Vary Excipient Type/Buffer Additives and Concentration (30 Working Days)
The low pH conditions of the elution buffer used during protein A purification is a potential
source of aggregation as described above. Based on an experiment conducted by A.A. Shukla et
al, sucrose, NaCl, and urea buffer additives will be tested for each mAb type (Shukla, 2007).
Variables tested in this experiment include additive type, pH, and additive concentration. The
objective of this experiment is to optimize elution buffer conditions to minimize aggregation
during low pH conditions of Protein A purification. This experiment is conducted last to utilize
pH and aggregation data found in the earlier experimental phases.
5. 4
Experimental Details
Phase 1
The problem at hand pertains to a specific mAb undergoing downstream process
development yet failing to maintain stability and readily aggregating in the drug substance. The
objective of phase 1 is to predict the inherent tendency of a mAb to aggregate. This analysis is
achieved with predictive simulations via a software called TANGO. TANGO observes the mAbs
known amino acid sequence and structure to identify aggregation-prone regions (APRs) in the
protein (Kant, et al., 2017). Additionally, it is important to note that TANGO distinguishes between
APRs that are protected in the thermodynamically stable hydrophobic core and those that are more
easily exposed (Kant, et al., 2017). While the stability issues under question pertain to a singular
mAb (mAb #1) undergoing downstream development, the versatility of this study is expanded by
the inclusion of two additional mAbs (mAb #2 and mAb #3) that are similar in function and
behavior yet have successfully established processing techniques. By the inclusion of these
reference mAbs, a robust analysis of mAb #1’s present issue may be provided in reference to the
specific amino acid sequence and the challenges it may pose. The results of this phase will be
compared to the results of phases 2-4. This will gauge the accuracy of TANGO and determine if
the predictive approach can be implemented in the initial selection process to eliminate
aggregation-prone mAbs, thus speeding up future research and development.
Phase 2
The objective of this phase is to determine when and how the majority of mAb aggregates
develop to focus on future mitigation efforts. A full factorial design is tabulated in Table 1. It
includes 24 experiment conditions, two control conditions, and three baseline conditions. The
experiment samples differ in two parameters: pH and mAb type. As discussed previously, three
different mAbs are evaluated. Meanwhile, the eight pH conditions that are evaluated are 3.0, 3.2,
3.4, 3.6, 3.8, 4.0, 4.5, and 5.0. These pH values are selected upon the consideration of standard
protein A elution conditions (pH 3-4). A pH of 4.5 and 5.0 observe alternative conditions that may
prove less harmful to mAbs during alternative forms of viral inactivation (e.g. via surfactants).
Meanwhile, the time-dependent nature of aggregation during the low pH hold is observed by the
sampling process over a 7-hour period. The 2 control conditions are formed with mAb #2 held at
6. 5
a moderate pH of 4.0 (samples #25 and #26) to measure data precision. The baselines, one of each
mAb at neutral pH, serve as the base values for SEC and DLS analysis.
Table 1. All variable conditions tested in phase 3 and control samples
Sample # pH mAb used
1 pH 3.0 #1
2 pH 3.0 #2
3 pH 3.0 #3
... .... ...
25 pH 4.0 #2
26 pH 4.0 #2
27 Neutral pH #1
28 Neutral pH #2
29 Neutral pH #3
To observe the mAb condition over the course of the hold, one sample will be collected
from the bulk at the start of the hold. This sample will be split into a triad and transferred to a
microplate-well, treated with ANS, and measured for its fluorescence with respect to time
(Wälchli, R, 2019). ANS is used to measure mAb degradation, showing modifications in surface
hydrophobicity associated with denaturation and aggregation (Wälchli, R. 2019). An increase in
ANS fluorescence shows an increase in protein denaturing. At the end of the 7-hour hold period,
the bulk solution will undergo titration to natural pH, another ANS sample is taken and split into
a triad. The fluorescence of these samples are to be observed for an additional 5 hours and
compared to that of the baseline neutral sample of the same mAb (Wälchli, R, 2019). Lower pHs
and longer holds are expected to yield higher fluorescence values. In Figure 1 below, the result of
a similar experiment shows an increase in ANS fluorescence over time at low pH.
7. 6
While ANS is underway, samples are extracted from each condition’s bulk once every 30
minutes for the first 2 hours of the low pH hold, which marks the typical time required for
inactivation. Additional samples are collected once every hour for 5 hours, a total of 7 testing
hours. Each sample is split into two tests, one at the hold-pH and on titrated back to the neutral
pH. Subsequent evaluation is achieved via Size Exclusion Chromatography (HP-SEC). At the end
of the hold, a sample is collected from the titrated bulk solution, allowed to stand for 1 hour, and
run through the HP-SEC. Peaks below the elution time of the mAb’s baseline indicate high
molecular weight aggregates and peaks above this baseline indicate low molecular weight
fragments. Figure 2 shows how SEC data corresponds to different aggregate sizes (Wang, 2018).
The amount of aggregation and fragmentation at low pH is compared to the results post-titration
to confirm or deny the proposed mechanism. If the aggregation rates are noticeably higher post-
titration, the focus of the experiment will turn towards preventing degradation at low pH.
Figure 1. ANS fluorescence
intensity of mAb 1 (a) and
mAb 2 (b) at different solution
pHs, over time. A lower pH
there is a sharper increase in
fluorescence. Adapted from
Ruben Wälchli et al.
Figure 2. Different
aggregate sizes
correspond to
different elution
times. The area
under the curve
corresponds to
particle abundance.
Adapted from
Shunhai Wang et al.
8. 7
Phase 3
The objective of this study is to couple phase 2 results with the observed pH dependence
of protein A elution performance and subsequent viral inactivation. With this interest in mind, each
of the tested eluent and hold conditions are generated with buffers, concentrations, and pHs that
are identical to those of the previous study. Hence, 8 Citrate/Phosphate elution buffers are prepared
with each differing only in their pH: 3.0, 3.2, 3.4, 3.6, 3.8, 4.0, 4.5, 5.0. These buffers serve in the
elution step of the Protein A and set the pH condition for subsequent viral inactivation.
Due to the complexity of potential contaminants faced in downstream processing, the
extent of this experiment is made uniform by the use of a purified mAb stock when formulating
the load-condition solution. This stock is one that is of known mAb type and concentration, free
of undesirable contaminants, and set to the standard loading condition of pH 7 with a
citrate/phosphate buffer. To test the system’s performance, clearly defined stocks of contaminants
are to be generated and spiked into the load-condition solution at known concentrations. One such
class of contaminant stocks will be high concentration HCP-DNA solutions. Three HCP-DNA
stock solutions will be generated; one stock solution per each mAb type, each utilizing the specific
DNA of the mAb’s associated host cell. Accounting for variation in the HCP-DNA among mAbs
enables robust analysis of the elution pH and its impact on eluted mAb purity. Furthermore, a
singular viral stock solution is generated with a mixture of 16 common viral strains, provided in
Table 2, in accordance with standard testing methods (Gombold et al., 2014). Each virus will be
obtained from stock seed solutions, incubated in fetal bovine serum, purified, and placed in a
neutral pH citrate/phosphate buffer. In conclusion, 3 contaminant-spiked neutral-pH load-
condition solutions are generated, varying only in the present mAb type: mAb 1, mAb 2, mAb 3.
Each load solution contains known concentrations of HCP-DNA, all 16 viral strains, and one of
the three mAbs.
9. 8
Table 2. Viral strains for system spiking and performance testing adapted from Gombold et al.
Viral Class Viral Strain Viral Class Viral Strain
Adenoviridae:
Adenovirus 5
Herpesviridae:
Simian
Cytomegalovirus
Adenovirus 41
Herpes Simplex Virus
Type 1
Polyomaviridae: Simian Virus 40
Paramyxoviridae:
Bovine Parainfluenza
Virus Type 3
Flaviviridae:
Bovine Viral Diarrhea
Virus (NY-1)
Coxsackie Virus A16
Orthomyxoviridae:
Influenza A Coxsackie Virus B3
Measles Rhabdoviridae:
Vesicular Stomatitis
Virus
Mumps Togaviridae: Rubella Virus
Picornaviridae: Echovirus 11 Picornaviridae: Rhinovirus 2
The experiment itself will be conducted as a factorial design, testing all possible
combinations of load solution mAb type and elution buffer pH; 24 experimental conditions. All
additional formulation parameters and column chromatography system parameters are to remain
constant at some preselected platform condition. Two additional control studies will be utilized in
an effort to evaluate the natural variance of the measurement techniques. MAb type #2 at elution
pH 4 will serve as the control and the experimental variance is set to the standard deviation of the
triad at this condition. For each test condition, a sample is collected from the load solution pre-
chromatography and from the eluent, low-pH hold, material immediately post-chromatography.
Following the experimental methods of Phase #2, samples are collected from the low-pH hold
every 30 minutes for the first 2 hours and every hour for an additional 5 hours.
All samples are to undergo three assays. The first of which is HP-SEC, discussed
previously, in the interest of evaluating the % purity pre and post chromatography in addition to
measuring the product yield. Furthermore, quantitative PCR (qPCR) will evaluate the presence of
residual HC-DNA in the post-elution samples as a measure of the system’s purification
performance (). This metric is provided in terms of the Cq, Quantitative Cycle, in which a larger
10. 9
value pertains to a lesser presence of HC-DNA and a greater quality of separation (Ichihara, 2018).
Lastly, all samples will receive a viral presence test with in vitro techniques. In this test, each
sample undergoes serial dilutions, creating a subsample array varying only in the dilution factor
(Gombold et al., 2014). Each subsample is split three ways and inoculated onto three distinct test-
cell plates; MRC-5, HeLa, and Vero (Gombold et al., 2014). Following the 14-day incubation and
analysis protocol set forth by Gombold et al., the presence of viral particles is deduced. Observing
the maximum dilution factor of observed virality, and the associated test-cell’s known limit of
detection, the viral concentration can be deduced. The cumulative data of phase 2 and phase 3 will
support the selection of an optimal pH condition.
Phase 4
The objective of Phase 4 experimentation is to optimize protein A elution buffer conditions
to minimize protein aggregation caused by low pH conditions. Three buffer additives will be tested
including sucrose, NaCl, and urea. These three buffer additives have shown to reduce aggregation
during low pH elution in previous studies (Shukla, 2007). The variables tested in this experiment
include additive type, additive concentration, pH, and mAb type.
A full factorial design will be used to assess each of the buffer condition variables. Each
additive will be tested at three concentrations for each mAb type at three well-performing pH
values found in phase 2 and 3. A base or platform buffer control will be tested at each of the pH
values. One trial will be conducted three times to assess the variability of the experiment. This
results in 32 trials per mAb for a total of 96 trials. Conducting a full factorial will enable the study
to include possible variable interactions. Trial condition variables are listed in Table 3 below.
11. 10
Table 3. All variable conditions of the full factorial design per mAb.
pH Buffer Additive Additive Concentration (M) mAb type
No. 1 Sucrose 0.5 mAb #1
No. 2 NaCl 1 mAb #2
No. 3 Urea 2 mAb #3
- Platform Buffer -
Percentage of high molecular weight species, HMW%, versus time will be measured for
each trial by SEC. This will indicate the fraction of the elute that is aggregated versus the stable
mAb fraction. An example of the SEC results are shown in Figure 3a below. HMW% versus
elution time will be plotted for each trial to analyze the trend in aggregated species. An example
of this graph is shown below in Figure 3b. The trial with the lowest HMW% trend over time will
indicate the best buffer conditions for each specific mAb type.
Figure 3. a) Absorbance units
versus time in minutes of a
protein A mAb elution adapted
from b) Percentage of HMW
species versus time in hours of
protein A elution fractions under
varying buffer conditions.
Adapted from A.A. Shukla et al.
12. 11
Each lab-scale protein A purification run is expected to be around 10 minutes. 10 additional
minutes will be needed for column cleaning between runs. Therefore, each trial is expected to
require 30 minutes to complete. This requires a total of 48 working hours for all 96 trials. Including
the time needed for data analysis and sample preparation, the total time for experiment phase 4
will be approximated to 30 working days. The results of this study will aid in both present and
future mAb aggregation mitigation efforts.
13. 12
References
1. A. A. Shukla et al. (2007) Protein Aggregation Kinetics during Protein A
Chromatography. Journal of Chromatography, 1171, 22-28.
https://doi.org/10.1016/j.chroma.2007.09.040
2. Gombold, J., et al. (2014) Systematic Evaluation of In Vitro and In Vivo Adventitious
Virus Assays for the Detection of Viral Contamination of Cell Banks and Biological
Products. Vaccine, 32(24), 2916-2926. https://doi.org/10.1016/j.vaccine.2014.02.021
3. Ichihara, T., ITO, T., Galipeau, K., Gillespie, C., (2018) Integrated Flow-Through
Purification for Therapeutic Monoclonal Antibodies Processing. MABS, 10(2), 325-334.
https://doi.org/10.1080/19420862.2017.1417717
4. Kant RVD, Karow-Zwick AR, Durme JV, et al. (2017) Prediction and Reduction of the
Aggregation of Monoclonal Antibodies. Journal of Molecular Biology, 429(8),1244-
1261. doi:10.1016/j.jmb.2017.03.014.
5. Vázquez-Rey, M., & Lang, D. A. (2011, July). Aggregates in Monoclonal Antibody
Manufacturing Processes. Biotechnology and Bioengineering, 108(7), 1494-1508. doi:
10.1002/bit.23155.
6. Wälchli, R., et al. (2019, December 11). Understanding mAb Aggregation During Low
pH Viral Inactivation and Subsequent Neutralization. Biotechnology and Bioengineering,
117(3). https://doi-org.proxy-um.researchport.umd.edu/10.1002/bit.27237
7. Wang, S., Liu, A. P., Yan, Y., Daly, T. J., & Li, N. (2018, March 16). Characterization of
Product-related Low Molecular Weight Impurities in Therapeutic Monoclonal Antibodies
Using Hydrophilic Interaction Chromatography Coupled with Mass Spectrometry.
Journal of Pharmaceutical and Biomedical Analysis, 154, 468-475. https://doi-org.proxy-
um.researchport.umd.edu/10.1016/j.jpba.2018.03.034